LLM-Guided Reinforcement Learning for Audio-Visual Speech Enhancement

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chen, Chih-Ning, Hou, Jen-Cheng, Wang, Hsin-Min, Chien, Shao-Yi, Tsao, Yu, Zeng, Fan-Gang
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908892986867712
author Chen, Chih-Ning
Hou, Jen-Cheng
Wang, Hsin-Min
Chien, Shao-Yi
Tsao, Yu
Zeng, Fan-Gang
author_facet Chen, Chih-Ning
Hou, Jen-Cheng
Wang, Hsin-Min
Chien, Shao-Yi
Tsao, Yu
Zeng, Fan-Gang
contents In existing Audio-Visual Speech Enhancement (AVSE) methods, objectives such as Scale-Invariant Signal-to-Noise Ratio (SI-SNR) and Mean Squared Error (MSE) are widely used; however, they often correlate poorly with perceptual quality and provide limited interpretability for optimization. This work proposes a reinforcement learning-based AVSE framework with a Large Language Model (LLM)-based interpretable reward model. An audio LLM generates natural language descriptions of enhanced speech, which are converted by a sentiment analysis model into a 1-5 rating score serving as the PPO reward for fine-tuning a pretrained AVSE model. Compared with scalar metrics, LLM-generated feedback is semantically rich and explicitly describes improvements in speech quality. Experiments on the 4th COG-MHEAR AVSE Challenge (AVSEC-4) dataset show that the proposed method outperforms a supervised baseline and a DNSMOS-based RL baseline in PESQ, STOI, neural quality metrics, and subjective listening tests.
format Preprint
id arxiv_https___arxiv_org_abs_2603_13952
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle LLM-Guided Reinforcement Learning for Audio-Visual Speech Enhancement
Chen, Chih-Ning
Hou, Jen-Cheng
Wang, Hsin-Min
Chien, Shao-Yi
Tsao, Yu
Zeng, Fan-Gang
Sound
Artificial Intelligence
Audio and Speech Processing
In existing Audio-Visual Speech Enhancement (AVSE) methods, objectives such as Scale-Invariant Signal-to-Noise Ratio (SI-SNR) and Mean Squared Error (MSE) are widely used; however, they often correlate poorly with perceptual quality and provide limited interpretability for optimization. This work proposes a reinforcement learning-based AVSE framework with a Large Language Model (LLM)-based interpretable reward model. An audio LLM generates natural language descriptions of enhanced speech, which are converted by a sentiment analysis model into a 1-5 rating score serving as the PPO reward for fine-tuning a pretrained AVSE model. Compared with scalar metrics, LLM-generated feedback is semantically rich and explicitly describes improvements in speech quality. Experiments on the 4th COG-MHEAR AVSE Challenge (AVSEC-4) dataset show that the proposed method outperforms a supervised baseline and a DNSMOS-based RL baseline in PESQ, STOI, neural quality metrics, and subjective listening tests.
title LLM-Guided Reinforcement Learning for Audio-Visual Speech Enhancement
topic Sound
Artificial Intelligence
Audio and Speech Processing
url https://arxiv.org/abs/2603.13952